This study was carried out to evaluate the quality of river water by analysis of land use in drainage basins and by estimating the N budgets. The drainage basins of Shibetsu River (Shibetsu area) and Bekkanbeushi River (Akkeshi area) in eastern Hokkaido, Japan were selected for a case study, and the evaluation of water quality was up-scaled to the regional level in Hokkaido by using the Arcview/GIS and statistical information. Water sampling was carried out in August 2001 and May 2002 in the Shibetsu and Akkeshi areas, respectively. The proportions of major land uses in drainage basins such as upland field, forest, urban area, wetland and wasteland for each sampling site were estimated by using topographic maps scaled at 1:25,000. The linear regression results showed that the correlation between NO3-N concentration and the proportion of upland in the drainage basins was highly and positively significant for both the Shibetsu area (r = 0.84, n = 57) and the Akkeshi area (r = 0.71, n = 73) at < 0.001 significance level. The regression coefficients or impact factors of river water quality were 0.015 and 0.0052 for the Shibetsu and Akkeshi areas, respectively. A comparison of these results with that of the previous study results in Hokkaido indicated that the impact factors were highest for intensive livestock farming areas (0.040), medium for mixed agriculture and livestock farming (0.020-0.030), and the lowest for grassland-based dairy cattle and horse farming areas (0.0052-0.015). The results of a simple regression analysis showed that the impact factors had a significant positive correlation with the cropland surplus N (r = 0.93, P < 0.01), chemical fertilizer N (r = 0.82, P < 0.05), and manure fertilizer N (r = 0.76, P < 0.05), which were estimated by using the N budget approach. Using the best-correlated regression model, impact factors for all cities, towns and villages of the Hokkaido region were estimated. The NO3-N concentrations for all major rivers in Hokkaido were predicted by multiplying the estimated impact factors by the proportion of uplands. The regression analysis indicated that the predicted NO3-N concentrations were significantly correlated (r = 0.62, P < 0.001, n = 203) with the measured NO3-N concentrations, reported previously. It can be concluded that estimating the proportions of upland fields in drainage basins, and calculating cropland surplus N enables us to predict river water quality with respect to NO3-N concentration.